Lambda isn't serverless...
..but it's really handy nonetheless.
It's a shame they called them lambda functions, because that means something to me already, but in AWS parlance, a lambda 'function' is a zipped up set of code and binaries (if you want) constituting a microservice that executes in a docker container on AWS' managed lambda infrastructure. Paired with Gateway API it's great. The really nice thing is that your code is invoked by events (e.g. an API call or reading off a queue), and, depending on the frequency of those events, AWS will provision multiple instances of this environment to ensure your events are soaked up as quickly as they arrive. When the events stop, then the number of environments quickly scales to zero. You pay purely for memory and CPU usage.
The 'lambda function' is typically node.js but you can then invoke anything at all you've uploaded - e.g. a golang binary to do the work, but most of the time, your microservice is just going to call another API (such as a dynamoDB or S3 or another queue).
So there you go - free yourself from EC2 cluster management, nginx configuration, etc and still do scalable stuff which costs nothing when there are no events coming in but reacts within expected thresholds when an event does appear..